| 注册
首页|期刊导航|新疆农业大学学报|基于iTransformer网络的富蕴县、福海县天然草原植被盖度反演方法研究

基于iTransformer网络的富蕴县、福海县天然草原植被盖度反演方法研究

田晶晶 吴音 卢宇航 刘聪 李英杰 尼露帕尔·阿迪力

新疆农业大学学报2025,Vol.47Issue(3):212-221,10.
新疆农业大学学报2025,Vol.47Issue(3):212-221,10.DOI:10.20088/j.cnki.jxau.2025.03.006

基于iTransformer网络的富蕴县、福海县天然草原植被盖度反演方法研究

Retrieval of Natural Grassland Vegetation Coverage Based on iTransformer Network in Fuyun and Fuhai Counties

田晶晶 1吴音 1卢宇航 2刘聪 2李英杰 2尼露帕尔·阿迪力2

作者信息

  • 1. 新疆乌鲁木齐市 自然资源规划研究院,乌鲁木齐 830011
  • 2. 北京数慧时空信息技术有限公司,北京 100070
  • 折叠

摘要

Abstract

In order to improve the inversion accuracy of natural grassland vegetation coverage under com-plex terrain conditions in Fuyun County and Fuhai County in northern Xinjiang's areas,this study intends to propose and evaluate a new inversion method based on the iTransformer network.Using GF-1 satellite imageries from July to September,2024 and field-measured vegetation coverage data,we conducted a com-parative study of three models for grassland coverage inversion:the deep learning iTransformer model,the machine learning random forest model,and the polynomial regression model,and then analyzed the impact of digital elevation model(DEM)terrain feature on the model accuracy.The result showed that the iTransformer model significantly outperformed random forest and polynomial regression models in both accuracy and generalization capability;Supplementing diverse terrain samples substantially enhanced iTransformer's generalization ability,reducing mean absolute error(MAE)and root mean square error(RMSE)by 12.2%and 13.1%,respectively;Incorporating DEM features further improved iTransformer's accuracy by 2%-9%.Through this method,we developed a well-performing grassland vegetation cover-age inversion model.For Fuyun County and Fuhai County,the coefficients of determination(R2)were both 0.83,while the error evaluation indicators,including MAE and RMSE,were 7.28%and 9.74%,and 6.89%and 8.87%,respectively.The prediction of grassland vegetation coverage distribution in study area was completed based on this model.

关键词

天然草原/植被盖度/iTransformer模型/GF-1影像/数字高程模型

Key words

natural grassland/grass coverage/iTransformer model/GF-1 images/digital elevation model

分类

农业科技

引用本文复制引用

田晶晶,吴音,卢宇航,刘聪,李英杰,尼露帕尔·阿迪力..基于iTransformer网络的富蕴县、福海县天然草原植被盖度反演方法研究[J].新疆农业大学学报,2025,47(3):212-221,10.

基金项目

新疆维吾尔自治区重点研发项目(2022B02005-2-3) (2022B02005-2-3)

新疆维吾尔自治区研究生科研创新计划项目(XJ2024G141) (XJ2024G141)

新疆农业大学学报

1007-8614

访问量0
|
下载量0
段落导航相关论文